Graduate Seminar
&
Student Technical Conference
Thursday, March 15, 2018
Department of Geodesy and Geomatics Engineering
University of New Brunswick
The Department would like to welcome you to the
Spring 2018 Graduate Seminar & Student Technical Conference
When:
Thursday, March 15, 2018 at 9:25 am
Where:
Head Hall – ADI Studio, Room C25
Department of Geodesy and Geomatics Engineering
Geodesy and Geomatics Engineering
Graduate Seminar and Student Technical Conference
Spring 2018
Chair: Heather Nicholson (MScE; year 3)
Thursday, March 15th 2018
9:25 Welcome note
9:30 Investigating the Use of Vessel-based GNSS Water Height Determination for
Validation of CVDCW HYVSEPS
Weston Renoud (MScE; year 6)
9:45 A RESTful API for what3words Extensions Encoding
Wen Jiang (MScE; year 2)
10:00 Effect of Glass Refraction on 3D Laser Scanning Measurement
Enuenweyoi Daniel Okunima (PhD; year 3)
10:15 Kinematic Precise Point Positioning with Map Constraints Performance
Analysis
Emerson Pereira Cavalheri (PhD; year 2)
10:30 A Local Projection for Integrating Geodetic and Terrestrial Coordinate
Systems
Mike Bremner (MScE; year 3)
10:45 Least-Squares Spectral Analysis of Zenith Total Delay Time Series
Omeiza A. Mayaki (MScE; year 3)
11:00 Break
11:15 Formulation and Validation of a Global Laterally Varying Topodensity
Model
Michael Sheng (PhD; year 4)
11:30 Precise Point Positioning Convergence Time Evolution and External Initial
Coordinate Integrity Analysis – Preliminary Results
Marco Mendonça (PhD; year 4)
11:45 Modelling Tropospheric Error for GNSS Reflectometry
Thalia Nikolaidou (PhD; year 4)
12:00 A DSM-Assisted Image Model for Improved Surface Reconstruction from
Satellite Images
Alexander Turner (MScE; year 2)
12:15 Improving the Accuracy of Extracting Surface Water Quality Levels using
Remote Sensing, Artificial Intelligence, and Water Quality Index
Essam Sharaf El Din (PhD; year 4)
12:30 RGBW DEMOSAICKING using a Collaborative Interpolation between
Panchromatic and Colour Pixels
Fatemeh Fathollahi (PhD; year 4)
12:45 Geometric Constraints for Improved Automatic Image Matching of Satellite
Images
Alexander Turner (MScE; year 2)
1:00 Closing Remarks
Spring 2018 Graduate Seminar & Student Technical Conference
ABSTRACTS
You may contact the Authors for a copy of the full papers.
Department of Geodesy and Geomatics Engineering
Investigating the Use of Vessel-based GNSS Water Height Determination for
Validation of CVDCW HYVSEPS
Weston Renoud
Email: [email protected]
Abstract
The Continuous Vertical Datum for Canadian Waters (CVDCW) project Hydrographic Vertical
Separation Surfaces (HyVSEPs) released in 2016 promises access to chart datum with the use of
GNSS positioning anywhere in Canadian waters. While accuracy targets of <10cm were believed
to be achieved in the development of the HyVSEPs, the approach relied on biasing to published
benchmarks and has yet to be validated.
The Ocean Mapping Group (OMG) has a history of mapping research in Canadian waters and an
interest in exploring the use of HyVSEPs for the reduction of bathymetric data in future work.
With previous work in global navigation satellite system (GNSS) based water height deter-
mination, OMG has existing data sets that may be viable for validation of the HyVSEPs.
This work investigates the use of existing OMG data holdings for use in validating HyVSEPs.
Specifically the use of data acquired on the RV Nuliajuk platform mobilized and operated by OMG
from 2012 to 2015 is explored. While equipped with multiple GNSS receivers, the previous
research focused on tidal model based reduction of bathymetric heights leaving the recorded raw
GNSS observations unexplored. This work looks at precise point positioning (PPP) processing of
5 days of GNSS data from 2013 and applying correctors for vessel dynamics to recover orthometric
water heights with the vessel.
The methods described in this research are shown to have a 0.003 meter (1𝜎=0.110 𝑚𝑒𝑡𝑒𝑟𝑠) mean
difference with observed tide gauge water heights which supports the use of RV Nuliajuk data
holdings for validating of the HyVSEPs in the eastern Arctic.
A RESTful API for what3words Extensions Encoding
Wen Jiang
Email: [email protected]
Abstract
With the advent of location-based services, the demand for location data has dramatically
increased. Geocoded locations have become necessary in many GIS analysis, cartography and
decision-making workflows. A reliable geocoding system that can effectively return any location
on earth with sufficient accuracy is desired. This study is motivated by a need for a geocoding
system to support university campus applications. Address-based geocoding systems have been
used for decades. However, they present limitations in address resources, address standardization
and address database maintenance. These limitations have recently sparked an interest in devel-
oping alternative geocoding systems that apply alphanumeric codes as a reference to locations,
such as Geohash, Google’s Open Location Code, and what3words to name a few. Comparing to
other geocoders, what3words (w3w) has many advantages. It uses a simple format of code con-
sisting of three words, it is less error-prone, codes are easier to memorize, and multiple languages
are supported. However, its fixed resolution (consisting of 3 meters by 3 meters square cells) and
lack of consideration of the third dimension may limit its applicability. To better sup-port
geographic applications with special requirements, the w3w geocoding system needs to be
extended. This paper proposes extensions of w3w in two aspects: variable-resolution and third
dimension support. And a HTTP-based RESTful API that implements QTEM extensions has been
developed to validate the geocoder processing in the two-way transformation between the
extended codes and coordinates and defining simple line and polygon features, as well as to pro-
vide web services for data interaction and information exchange between client programs. This
research could support the need of a university campus’ facility management, emergency evac-
uation and route navigation planning, student survey data management, and other location based
services.
Effect of Glass Refraction on 3D Laser Scanning Measurement
Enuenweyoi Daniel Okunima
Email: [email protected]
Abstract
3D laser scanning is a non-contact technology that allows rapid collection of three dimensional
(3D) data of objects and surfaces. This technology provides solutions to several industries
including transportation, cultural heritage, topographic surveying, volume estimations, reverse
engineering, and as-built surveying of facilities and buildings.
Glass is an important material in the construction of different kind of buildings because it improves
aesthetics of buildings and provides a means for natural light to get into buildings. Hence it is
common to encounter glass during laser scanning documentation of buildings. In some instances,
laser scanning practitioners find it more convenient to scan through glass during the data
acquisition. In other instances, scanning though glass may be the only way to capture the object(s)
of interest with a 3D laser scanner when there is limited room to setup the scanner. Therefore, it is
important to determine the effect of glass refraction on the laser scan measurement.
Trimble TX5 3D laser scanner was used to scan three black and white checkerboard targets by
positioning the equipment so that the laser pulse from the scanner passes through either one or two
5 mm thick glass sheets before getting to the checkerboard targets. Two sets of laser scans
comprising four scans per set were obtained from two instrument heights to produce eight point
cloud dataset. For each set, the checkerboard targets were scanned without a glass sheet, with a
glass sheet in the front position (i.e. closer to the laser scanner), with a glass sheet in the rear
position (i.e. closer to the targets), and with both glass sheets in place. The results obtained shows
that the traveling of the laser pulse through glass increased the measured distance for both
instrument heights. The increase ranged from 3 mm to 6 mm. The effect of refraction on the
measurement was greater when the laser pulse passes through both glass sheets.
Kinematic Precise Point Positioning with Map Constraints Performance
Analysis
Emerson Pereira Cavalheri
Email: [email protected]
Abstract
For high precision kinematic applications using the Precise Point Positioning (PPP) technique the
use of the precise carrier-phase observable is imperative. However, obstructions in the line of sight
in challenging environments, e.g. urban space, limits the carrier-phase use. Attempts of providing
an a-priori information to the PPP filter has been developed. The main success was only in the
reduction of the initial convergence time, however the reconvergences in the middle of the
kinematic collection, to the best of our knowledge, has not been yet addressed. In this contribution,
an alternative a-priori position coming from a ground truth trajectory is used on a kinematic PPP
collection at every epoch. To connect where the rover is in this trajectory, a map-matching
algorithm is necessary. At every epoch, using a single point position (SPP) and uncertainty coming
from the less accurate pseudoranges, the map-matching localizes the receiver in the trajectory and
determines a search space where the true position will likely to be. Finally a search of the best
candidate is performed using the Kalman filter residuals. The results have shown that
convergences and re-convergences could be eliminated, and when compared to the PPP-only
solution, the pseudorange ionosphere-free residuals have shown an improvement of 17.5 cm and
24.3 cm in the residuals bias and standard deviation, as for the carrier-phase ionosphere-free
residuals, both solutions were unbiased, and the standard deviation were 3 times better than the
PPP only solution residuals.
A Local Projection for Integrating Geodetic and Terrestrial Coordinate
Systems
Mike Bremner
Email: [email protected]
Abstract
Within the larger research project, a need was found for a system capable of transforming between
global geodetic coordinates and local coordinates defined by terrestrial measurements. To fulfil
this need, a system is proposed that uses an extension of the Double Stereographic Map Projection
for transformations and a parametric least squares estimation to calculate the parameters for the
map projection.
Research was performed on existing solutions; however, literature was found to be sparse.
Research was also performed on map projections, with particular interest on the Double
Stereographic Projection.
The transformation equations of the Double Stereographic Projection are extended to include a
rotation parameter, and a least squares estimation for the calculation of projection parameters is
devised. The original Double Stereographic Projection and the extended model are both
implemented, as is the projection parameter estimation.
Tests performed comparing the results of the two map projection implementations are found to be
successful. A test comparing the results of the projection parameter estimation against local grid
transformation with known parameters is also successful. Future tests are also proposed, although
some aspects are expected to be indeterminable.
Least-Squares Spectral Analysis of Zenith Total Delay Time Series
Omeiza A. Mayaki
Email: [email protected]
Abstract
As a product of ground-based Global Navigational Satellite Systems (GNSS) observations, the
derived Zenith Total Delay (ZTD) plays an important role in meteorological studies. Several
institutions undertaking atmospheric research utilize networks of GNSS continuously operating
reference stations (CORS) as additional source of ZTD information which they assimilate
operationally into regional and global Numerical Weather Prediction Models (NWP/NMW) for
accurate weather forecasting and now-casting. In West Africa, there are no operational GNSS
networks in which its ZTD data are assimilated into NWP models, or used for GNSS meteorology
generally. In Nigeria however, the suitability of the Nigerian GNSS reference Network (NIGNET)
for meteorological applications is being investigated. Analysis of a collection of ZTD time series
can provide insight about the seasonal variability of the weather in the different climatic regions
of the country. This report presents a study of the temporal variation of the periodic constituents
of ZTD time series computed from NIGNET data sets in the precise point positioning (PPP) and
from NWM, using Least Squares Spectral Analysis (LSSA). Results show strong periodic
constituents that relate to known seasonal weather phenomena in Nigeria, below and at the 1-year
time span, and constituents with periods longer than a year. This analysis enables the
characterization of the nature of the troposphere over Nigeria in both the local and regional sense.
It also further affirms the prospective relevance of ground-based GNSS as a tool for future
exploration of Nigerian weather and climate.
Formulation and Validation of a Global Laterally Varying Topodensity Model
Michael Sheng
Email: [email protected]
Abstract
The distribution of masses within topography is extremely important for evaluation of a rigorous
geoid model. Present day models of topographical density are either determined via gravity field
inversions or are at too coarse a resolution to be useful on a regional level. The ideal solution
would be to formulate a 3-dimensional (3D) density model but such a dataset does not currently
exist and would likely be extremely complicated and expensive to develop. As a substitute, a 2-
dimensional (2D) model (UNB_TopoDens) is presented in this paper through the application of a
lithological map. The methodology that was applied in the formulation of the UNB_TopoDens
model is discussed and then this model is validated against present-day global and regional models
in both its spatial and spectral forms. Apart from some water and ice covered regions, the
UNB_TopoDens model validates very well with all regional and global models with the
differences all falling within the 95% confidence region associated with the UNB_Topodens
model. UNB_TopoDens is the first laterally varying topographical density model with associated
error estimates that spans the globe at a 30” x 30” resolution and is there-fore extremely useful
across various disciplines.
Precise Point Positioning Convergence Time Evolution and External Initial
Coordinate Integrity Analysis – Preliminary Results
Marco Mendonça
Email: [email protected]
Abstract
Seamless positioning is a recurring challenge among the scientific community. Modern
technologies allow developers to work with multiple signals in parallel and solve systems of
equations that a decade ago would not be handled in real-time. In this context, the idea of
integrating satellite positioning techniques, such as Precise Point Positioning (PPP), with other
techniques is facing an upsurge. Some of the techniques that are considered candidates to be
integrated with satellite positioning are becoming more and more convenient when PPP faces its
limitations. In urban-canyons, for example, kinematic PPP may not provide information reliable
enough for most of applications. From autonomous vehicles, requiring positions better than 50
cm, to pedestrian navigation, at the decameter level, GNSS errors may cost time, money and even
cause accidents. On the other hand, where GNSS may face most of its limitations, other positioning
sources may thrive. Under such circumstances, the signal-of-opportunity, hereafter SOOP,
presents itself as a potential source of information to fill the limitations of PPP platforms. By
utilizing signals that are not specifically design for positioning, yet provide spatially variable
information, one may integrate information originated from these systems with the PPP algorithm
in order to increase its availability and reliability. In this study, a simulation of how the PPP
convergence time may improve in the presence of external information being used as initial
coordinates is performed. SOOP sources will provide coordinates in several different quality
levels, hence, a simulation of 500 different combinations of offset and standard deviations is
applied to the initial coordinates of a kinematic PPP processing and the results analyzed. This study
concludes that the combination of offsets and standard deviations of the input coordinates provides
essential information to the PPP algorithm, being able to significantly reduce the convergence time
in a kinematic processing.
Modelling Tropospheric Error for GNSS Reflectometry
Thalia Nikolaidou
Email: [email protected]
Abstract
In the recent years Global Navigation Satellite Systems (GNSS) signals have been exploited for
coastal sea lea level altimetry. The existing infrastructure along their low maintenance cost renders
to signals of opportunity for GNSS multipath reflectometry (MR). Latter studies have shown that
the error induced by the lower atmosphere causes a significant height/depth variance that if not
accounted properly can lead to underestimation of the tidal amplitudes. Two empirical models
that emerged to account for that, are based on the concepts of either angular or linear estimation
of the delay and are functions of the elevation angle of the satellite and the sea level height/depth.
The first model represents the geometric effect of the tropospheric error while the second attempts
to give an estimation of the speed retardation of the signal. Although these models are able to
express the tropospheric error to some extent, their rather simple formulation lacks a wholistic
modelling of the total error, leading to biased results up to meter level. We have implemented a
ray-tracing procedure to calculate the total error in a complete – three point boundary value-
schema: broadcasting satellite, reflecting surface, and receiving antenna. Moreover, we have
analyzed the models to gain insight about the dependence of delay on surrogate variables, such as
bending angle and mean refractivity, as well as independent variables, such as satellite elevation
angle, station altitude, and reflector height/depth. We present the results of the analysis and access
their accuracy against the ray-tracing simulations for a station located in an exceptional high
altitude by the shore. The use of the empirical models shows systematic errors that can reach
centimeter to decimeter level errors in tropospheric modelling which in turn can lead to meter level
error in the sea level height/depth estimation.
A DSM-Assisted Image Model for Improved Surface Reconstruction from
Satellite Images
Alexander P.R. Turner
Email: [email protected]
Abstract
Satellite imagery is of great appeal to engineers and environmental scientists alike for its low repeat
time and relatively low cost. However, satellite imagery is less straight forward for
photogrammetry than aerial frame cameras being that they are linear array cameras. This means
that instead of having a single principle point they have a different principle point for each line of
pixels, so one could say that the images have a principle line. The exterior orientation parameters
describing this line are, more often than not, not provided by the vendors but rather a series of
rational polynomial co-efficients (RPCs) that can be applied to a rational function model to convert
between image and object space. As with any measurement or calculated value, these RPCs
contain errors which can lead to false matching and false geometry. Several publications have
outlined that these errors can be reduced by applying an affine correction to the RPCs but do not
provide an automated method for doing so. This proposal outlines a methodology for an automated
RPC correction in such a way that false matching detection and surface reconstruction are
completed as bi-products of the operations.
Improving the Accuracy of Extracting Surface Water Quality Levels using
Remote Sensing, Artificial Intelligence, and Water Quality Index
Essam Sharaf El Din
Email: [email protected]
Abstract
Extracting accurate surface water quality levels (SWQLs) always presents a great challenge to
researchers. Existing methods of assessing surface water quality can only provide individual
concentrations of water sampling stations without providing the accurate SWQLs. Therefore, the
results of existing methods are usually difficult to be understood by decision-makers. On the other
hand, the water quality index (WQI) can simplify surface water quality assessment process to be
accessible to decision-makers. However, in most cases, the WQI reflects inaccurate SWQLs due
to the lack of representative water samples. It is very challenging to provide representative water
samples because this process is costly, labour intensive, and time consuming. To solve this
problem, a cost-effective method, which combines Landsat-8 imagery and artificial intelligence to
develop models to derive representative water samples by correlating concentrations of ground
truth water samples to satellite spectral information, is introduced. Our method was validated and
the correlation between concentrations of ground truth water samples and predicted concentrations
from the developed models reached a high level of coefficient of determination (R2) > 0.80, which
is trustworthy. Afterwards, the predicted concentrations over each pixel of the study area were
used as an input to the WQI developed by the Canadian Council of Ministers of the Environment
to extract accurate SWQLs, for drinking purposes, in the Saint John River, as the testing water
body. The results indicated that the SWQL was observed as 67 (Fair) and 59 (Marginal) for the
lower and middle basins of the river, respectively. These findings demonstrate the potential of
using our approach in surface water quality management.
RGBW DEMOSAICKING using a Collaborative Interpolation between
Panchromatic and Colour Pixels
Fatemeh Fathollahi
Email: [email protected]
Abstract
A newer generation of the standard colour filter arrays (CFA) contains panchromatic pixels along
with red (R), green (G), and blue (B) pixels. This CFA, known as RGBW, improves the SNR
compared to the standard CFAs since the sensitivity of the panchromatic sensor is higher than the
RGB sensors. However, the results of the demosaicking algorithms proposed so far still are not
satisfying and suffer from the severe artificial colours especially in areas of high-frequency
information.
In this paper, we propose a new procedure to improve RGBW demosaicking performance. This
algorithm involves a collaborative interpolation between panchromatic and colour pixels, instead
of a separate independent interpolation as performed in most demosaicking algorithms. Through
collaborative interpolation we incorporate the available colour pixels in the panchromatic
interpolation and vice versa, we incorporate the available panchromatic pixels in interpolating each
colour channel. Finally, the interpolated panchromatic and colour images are fused together
through UNB-pansharp technique. In this paper, we focus on Kodak Truesense pattern since a built
sensor with such pattern is available within BOBCAT (IGV-B2320) camera.
To evaluate the proposed method, we used the Kodak benchmark dataset and we followed the
same evaluation routine in literature to compare our results with other RGBW demosaicking
methods. Furthermore, we test our method using some real-world images taken by the Kodak
Truesense sensor. The quantitate evaluations in both test scenarios show that our proposed
algorithm outperforms other demosaicking methods as well as the camera’s result, especially in
terms of sensitivity and produces less artificial colours.
Geometric Constraints for Improved Automatic Image Matching of Satellite
Images
Alexander P. R. Turner
Email: [email protected]
Abstract
The prospect of using satellite imagery for engineering and environmental applications is an
exciting enterprise in the modern world of digital imagery and high end computer processing
capabilities. If the imagery is to be useful for automatic surface reconstruction or change detection
then it is necessary to have the capability to accurately match the points from one image to the
other. To limit the search area for matching points between two images Schenk (1999) outlines
two methods, vertical line locus and epipolar geometry. Since a scene with no surveyed ground
control points can only have a relative bias compensation based on approximate heights there is
an error in the model that is associated with the error in the approximate heights. This error is in
the direction of the images epipolar lines and so it will have no effect on the results derived from
epipolar geometry. The unique problems with satellite images, as opposed to terrestrial and aerial
frame cameras, is that the images are related to object space by means of a rational function model
which contain errors and that epipolar lines are parabolic in shape rather than the straight epipolar
lines found in frame cameras. Evidence is given to proven the non-straight nature of the epipolarity
of linear array cameras and the approximate known height variance that is needed for the best
approximation. This paper presents methodology on how a low resolution Digital Surface Model
(DSM) can be used to constrain epipolar segments to reduce the potential of falsely matched points.
University of New Brunswick Department of Geodesy and Geomatics Engineering
Head Hall – 15 Dineen Drive
PO Box 4400 Fredericton, NB Canada E3B 5A3